Hugging Face's logo

Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
Back to feed
mohit67890 
posted an update 21 days ago
Post
4,991
🥇 Imajev-4b is #1 of 50 on Image JevBench (v0.1.4, 29 Sep 2026), the leaderboard for AI models that make decisions from images.

A small demo built on imajev-4b: a closet stylist 👗
Request you to star it here so we can make it better - https://github.com/mohit67890/imajev.

Tap a piece and it reads the photo (red 75%, checked 99%), then the app picks bottoms, shoes and a bag from your own closet in the colours you like. Change your colours and the outfit changes.

Under the hood it's one request with one photo and 4 typed questions. Every option gets a probability, so the app applies its rules (one pattern per outfit) and ranks what's left. About 1.1 s per outfit on a Mac (MLX). Every % in the video is the model's real answer.

Also, thanks to @zenmagnets for the FP8 version for Blackwell GPUs: same calibration, 0.1 pt less accuracy on all 23,900 DecisionBench rows, ~1.6× faster 🙏
zenmagnets/Imajev-4B-FP8-SM120

🧠 Weights: mohit67890/imajev-4b
🚀 Demo: mohit67890/imajev
📊 Leaderboard: https://benchmarkheaven.com/image-jev-bench
yano2mch

I wonder how useful this would be for me.

Doing a few different projects.

  1. Generating images for cards that lack an image (RP cards, think Character Tavern, Character Hub or SillyTavern), and a lot of images have messups (like 10%-20%, the usual, screwed up hands, perspective where a bike is behind them 20 feet yet their hand is resting on the bike, extra limbs, etc) enough to qualify a regen. Wonder if this would do for finding those. Likely with the description maybe also an answer between 1-10 of how much it follows the prompt.

  2. Manga optimization. I need to identify color, black/white, dithered and grayscale. I have optimized scripts to get really good compression for each case. But if they are in the wrong set they either take up a lot more space, or get really bad downsampling results. A handful of cases the page is just text, so identifying that to OCR instead of keeping separate images.

  3. Text alignment. I got a couple OCR methods where paragraphs don't always get newlines, which needs done beforehand. Could i use it to compare against the OCR to identify which line(s) in the input should have extra spacing added?
In this post